Situation Element Extraction Based on Fuzzy Rough Set and Combination Classifier

Algorithm 1

Based on conditional attribute similarity and improved fuzzy rough set attribute selection algorithm.

	Input: network security information decision system , initial clustering threshold , and empty attribute set .
	Output: attribute reduction set that meets the requirements.
	Step 1: according to the k-order weighted average method, the lower approximation of fuzzy rough set under condition feature set is obtained, and then the dependence of decision feature on is obtained;
	Step 2: calculate the similarity matrix of condition attributes according to the similarity measurement formula of condition attributes;
	Step 3: complete the clustering of conditional attributes according to the direct clustering method and clustering threshold , and divide the set of conditional attributes into similar subsets ;
	Step 4: according to the maximum similarity criterion, select a representative condition attribute from each attribute similarity subset to form the secondary reduction attribute set ;
	Step 5: get the lower approximation of the fuzzy rough set under the attribute subset according to the method of order weighted average in the next neighborhood, and then get the dependence degree of the decision attribute on the attribute subset ;
	Step 6: judge whether depends on more than or equal to . If it does not meet the requirements, increase the clustering threshold and return to Step 3 to continue. Otherwise, complete the secondary reduction of conditional attributes and get the attribute set;
	Step 7: take attribute subset as the attribute set to be reduced, and select an unselected condition attribute from it to copy it to attribute subset ;
	Step 8: the lower approximation of fuzzy rough set under attribute subset is obtained according to the k-order weighted average lower approximation calculation method in the nearest neighbor region, and then the dependence degree of decision attribute on attribute subset is obtained, and then attribute is deleted from attribute subset ;
	Step 9: repeat Step 7 and Step 8 until every element in is traversed;
	Step 10: select the attribute that increases the dependence of decision attribute on attribute subset the most and add it to attribute subset and delete it from attribute subset ;
	Step 11: check whether the dependency of on is greater than or equal to that of after attribute is added to attribute subset . If it does not meet the requirements, it will return to Step 7 to continue. Otherwise, it will complete the final reduction of conditional attributes and output the final attribute reduction set .